how to deploy deep learning prototype to FPGA using HDL Coder?

Hi i have FPGA platform and like to play around with Matlab deep learning (like MatConvNet), then deploy on my FPGA platform for testing (both SW and HW), i know there are tools such as Embedded Coder and HDL Coder might help me on this, but not sure how and the limitation? please kindly share your input, thanks a lot.

Regards, Owen

[Alexander Vorobyev]

Example hdl code generation from CNN https://www.youtube.com/watch?v=dSWkawpTiM4

[owen chang]

thanks. but unfortunately i can't understand this language...

hopefully will have engilish version and more detail can be shared from there.

thanks for shaing

Owen

Answer 1

[Girish Venkataramani]

Hello Owen,

Apart from MatConvNet, MATLAB now supports deep learning natively in Neural Network Toolbox. See this for more info: https://www.mathworks.com/campaigns/products/ppc/google/deep-learning-with-matlab.html

For code-generation, we now support Deep learning networks to CUDA code with the new GPU Coder product that was launched in R2017b. We are currently actively working on adding code-generation support also for CPUs (C code) and FPGAs (HDL code) from these neural network objects that represent deep learning. You'll hear more on this in the next year.